Working with freely available data-sets and using machine-learning techniques, young ecologist Vijay Ramesh and his collaborators have shown that at least 17 bird species in the Western Ghats face graver risks of extinction, and their survival is more threatened, than previously held to be. The results of this study were published on April 25 in the journal Biological Conservation, a leading international journal of conservation biology published by Elsevier, the Dutch publishing house which also publishes prestigious journals like The Lancet, Cell, and ScienceDirect.

Firstly, why is this data important? That’s because the conservation status of any animal or plant species is a hugely significant ecological parameter, which is used to determine efforts that go into eliminating risks to the concerned species, as well as accurately assess the ecological threats facing the habitat.

Ramesh, a spatial and computational ecologist at the Cary Institute of Ecosystem Studies, and his colleagues found that range maps used by the International Union for Conservation of Nature (IUCN) fall short of protecting birds endemic to the biodiversity hotspot of the Western Ghats. When current range maps used by the IUCN, supplied by BirdLife International (BLI), a global partnership of conservation organisations focused on bird species, were assessed using modeling techniques that incorporate data on species sightings, land cover, and climate, 17 out of 18 species included in the study were found to be inaccurate and overestimated. This directly impacts the conservation status of these species.

“Of the 18 species in our study, 17 of these birds’ ranges have been severely overestimated by BLI and IUCN,” says Ramesh. This data directly impacts conservation efforts. “Moreover, we found that half of these species are not actually found in over 60% of the areas mapped by BLI. For instance, the Nilgiri Pipit, which is currently listed as ‘Vulnerable’ by the IUCN, has a range of less than 1,392 sq km, whereas BLI lists its range as 11,558 sq km, thereby overestimating the range by a staggering 88%, thus moving its threat status to ‘Endangered.’”

Other examples include the Nilgiri Flycatcher, whose range has been overestimate by 63%, the Wayanad Laughing Thrush, whose range is approximately 24,000 sq km whereas BLI lists its range as approximately 154,000 sq.km.

In fact, many species that are currently listed as “Least Concern” actually need to be listed in the threatened categories, says Ramesh. For example, the Rufous Babbler has a range of ~35,000 sq km but BLI lists its range as ~178,000 sq km. “That is a huge discrepancy, as you can imagine,” he adds.

How this discrepancy was unearthed is a fascinating story of using freely available data, and (in most cases) freely available technology, to create a new way of looking at the fragile ecology of the Western Ghats.

The group’s algorithmic models accurately plot the birds’ ranges and also take into account climate and habitat data, and land cover, which can vary tremendously over a given habitat. Ramesh feels that current methods don’t take into account all these data sets, and also fail to tap into the wealth of citizen science data on where birds are actually seen. “As a result, unsuitable habitat is often included in the species’ range, inflating the apparent extent of where it can be found,” says Ramesh.

The project started with field work done by one of the co-authors in the study, Sahas Barve, a postdoctoral associate in the Walters lab at Old Dominion University, Virginia, USA. He was in the field during his Master’s research years ago and his initial realization was that accurate range maps for most birds in the Western Ghats were rare.

The group — which includes, apart from Ramesh and Barve, Trisha Gopalakrishna, a tropical ecologist currently working at The Nature Conservancy (US), and Don Melnick, a professor of conservation biology at Columbia University — initially planned to publish a study with accurate ranges for 18 birds that they chose. However, when the researchers looked up IUCN’s range maps and visualized these maps on Google Earth, they saw that most of these maps for these birds had major townships, settlements, agricultural plantations within their range. They realised this could be skewing the range data for these bird species, and impacting their conservation status.

Then they obtained curated and filtered data from the world’s largest citizen science database, eBird. This data is growing exponentially. Incidentally, India is currently one of the top 4 contributors to this data set, which shows that India’s amateur bird watchers and conservationists are a dedicated, hardworking bunch. “To give you an idea of the size of the files that I dealt with, I worked through a dataset that is currently at 138 gigabytes,” says Ramesh.

Secondly, they acquired open source data on variables such as temperature and precipitation. Third, they used data from remotely sensed satellite imagery, which gave them an idea of the entire Western Ghats landscape. They also obtained information from the Indian Institute of Remote Sensing on land-cover at the scale of 23.5 metres. Ramesh is particularly excited about this freely available data-set. “This is one of the finest resolutions you can get for a freely available source, apart from the Sentinel data from the European space agency, where data is at a resolution of 10 metres,” he says.

Then the researchers used all of this information in a species distribution modeling framework (a species distribution model produces a robust distribution by estimating the relationship between the known location of a species and the environmental conditions that are characteristic of that location) and created a machine learning model that would make sense of the several layers of data to deliver the desired results.

“Machine learning is a relatively new field in ecology. Essentially, you provide information that is used to “train” a model — in our case, we are providing environmental information on where these birds have actually been recorded. This information will be used by the model to search across the entire Western Ghats landscape to find similar locations where such environmental information exists. We did this for each species,” explains Ramesh.

The machine learning approach the researchers used is called a Boosted Regression Tree (BRT). Ask Ramesh about the other technology, however ‘basic’, they used, and he rattles off: “Google Earth, ArcGIS (for spatial visualization and partial analysis), Python (programming language that has proved to be extremely useful in batch-processing of data, and saves a ton of time) and most importantly, the R programming environment for all the machine learning models that were built (which I have spent days on end trying to figure out a simple error).” He points out that all of this software is freely available except for ArcGIS, although there is an alternative which is called QGIS.

Technology plays a very important role in conservation today, say Ramesh and his cohorts. “Starting with a simple compass that has been used since time immemorial, to using apps to identify a bird species or record a bird song, technology has become an integral part of science and conservation today,” he says. He gives some examples of the best work: real-time monitoring of deforestation using satellite imagery by NASA where you can see images almost every 8 hours over a location to see which region might be deforested next.

On a different end of the spectrum is conservation genetics and Don Melnick, the fourth author in this study has been working in this field for over four decades. By taking a toe snip from a frog or by taking less than 1ml of blood from a bird species, one can identify if these populations are endangered due to reasons such as lack of gene flow, loss of diversity over time. “The famous case being cheetahs, whose genetic diversity is so low today, that almost every single cheetah in the wild is related to one another,” says Ramesh.

Their study should go some way towards ensuring the same fate is not met by bird species in the Ghats.

---